Differential cryptanalysis is a form of cryptanalysis which studies cryptographic algorithms by observing how differences in input affect differences in output.

In practice, this is usually a chosen-plaintext attack with a quite large number of plaintexts (but for "broken" algorithms still smaller than the key space).

The terminology is presented by Biham and Shamir and within IBM, the attack was formerly known as the "T-attack" or ‚ÄĚtickle attack.